Thanks to this question I became aware of the fact that we have (at least) these overlapping tags:
probability
probability-and-stats
statistics
Are "probability" and "statistics" sufficiently differentiated to keep each of those tags and eliminate probability-and-stats or should all of these ...
@YungHummmma lf your "plot" is empty and all you want are axes and ticks simply turn them on, e.g. Graphics[Circle[], Axes -> True]. If you have also a function to plot use either Prolog or Epilog to place the primitives behind or in front of the plot respectively, e.g. Plot[Sinc[x], {x, 0, 10}, Epilog -> Disk[{4, 1/2}, {0.5, 0.2}]]
I am trying to work out a better solution to mathematica.stackexchange.com/q/88085/121 and I am wondering if this can be done with PermutationGroup and/or related functionality. Do you have experience in that area?
@Mr.Wizard No, I would need to read it carefully to understand all implications. I used PremutationGroup in different situations, but after 19h awake there is no way I can say anything useful now.
Sorry.
But the questions sounds nice and has already some good answers afaics
I flagged this answer because I mis-read it and thought it was a thank-you answer and not a real one. I know my flag will be rejected and rightly so, but I wish I could have retracted it myself. The close-vote pop-up allows retraction, I request the flagging pop-up be given the same feature.
It seems version 10.2 has been released and it is another paid upgrade for Home Edition users. Am I the only one bothered by the idea of a paid upgrade only four months after the last one?
@Mr.Wizard Ar you bothered of it as someone paying for it, or as a potential reflection of the financial situation at WRI?
I didn't really spend a split nanosecond thinking of if I should make the purchase when it became available, although the change in frequency of paid upgrades might induce some thoughts.
Hypothetical: if through some fluke, you find out tomorrow that they'll be releasing a version with even more fixed bugs and less doc typos in three months for the same price, would you still feel okay about what you just paid for?
But to a large extent, I don't believe pricing depends at all on what gets delivered, but more of what kind of sums people categorize as "large" or "small", and what's their general brand recognition of things.
@Guesswhoitis. I'd probably hope they'd offer an subscription model with more clarity in update schedule and annual costs, and a meaningfully priced way to change licensing plans.
I don't actually consider annual cost of 500-700 euros for having always the latest version of Mma that bad, assuming bugs are fixed and useful new features introduced. A reasonably priced option to switch to non-upgrade path without annual costs should exist too, though.
Yep, there are dangers of that. And interesting vendor lock-in of demanding continuously updated version to be able to use examples posted in Mma.SE, or such.
For a 25-year-old system containing practically everything, Mma has done pretty well in my opinion.
I wonder if SW really had insight into naming of things and sufficiently non-ambiguous syntax from the start, or if it was lucky guess? I suppose the preceding SMP had shown that some approaches didn't scale...
Maybe he really was sufficiently megalomaniac to over-design the approach on naming and syntax with a 25-year plan, as he has sometimes claimed.
@Guesswhoitis. That is one of the things that bothers me: it is not unreasonable to pay for new features but bug fixes should be free. Where are the support releases like 10.0.3 and 10.1.1?
@Mr.Wizard Then again, I think it's unreasonable to wait for some new features for years just because they want an impressive list of new features. (Of course large new subsystems may take years to develop.)
@kirma Generous members of this community exclusively paid for my last two upgrades. However as a matter of principle I am concerned that much-needed bug-fix (support) releases are being replaced with pay-us-more upgrades that always contain new bugs in the ever-increasing functionspace.
@kirma The system is already perceived as aloof (for lack of a better term at the moment) for being a closed commercial system, by those using free alternatives. I think the never-ending cost of a subscription model would only further this problematic image.
But also: is there actual competition on the market? Especially from integrated free alternatives? I think this WRI behavior is possible mostly because there is no serious general competition, especially from free/open source offerings.
Their largest risk is not that profitable customers would be wooed to a competing product, but that they'd just stop paying for it.
Apart from some fields that go over my head, my use of Mma is all over the feature set, and there's always something that could be better or have nicer abstractions. I guess that's not necessary the case in many non-recreational use cases.
@kirma I think that people who don't already use Mathematica won't readily see that. When WRI sends people to present Mathematica at universities, the presentations are not well done or convincing either. They only focus on "cool stuff" such as tweeting and the only people who'd be convinced are those who don't / wouldn't use Mathematica on a daily basis to do their own anyway.
It might be true that the decision makers are not those who actually seriously use the product though ...
@Mr.Wizard The versioning model and the complete lack of understanding that you should not charge your precious users for bug-fix releases is only one of the things WRI does wrong. It starts by publishing a version 10.0.0 which is not only full of bugs, but where many of the new shiny features were just not ready to be published. And then in each minor release some bugs are fixed, new bugs are introduced and the unfinished features mature step by step until they finally (probably) can be used.
3
The suggestion for users is clear: Wait until the major version is mature enough so that it can be used. What good does it do when you have to work with a version that constantly crashes?
Funny story: I got an invitation to rate the WRI-Support for an issue I had reported. The issue was the WDX exporting bug that I posted here. The only answer I got from the support was that it is known and they are working on it. This means for me, someone already reported it and I wasted my time due to the lack of a public issue-database.
All other information, most importantly that WDX is very likely to be abandoned soon and that one should stick to MX if possible, and that Dataset probably won't never be supported natively, were shared by developers of WRI in their freetime by answering/commenting.
@Szabolcs Yes... but it is not the only company in the world that somehow fails to advertise their good technology to people that would be their real customers.
@halirutan How do you verify a version is "mature enough" ? Sadly "tested by customers" tends to be the most cost-effective method for non-critical systems.
Well, OK, regressions should be caught, of course... but producing good corner case test coverage for new functionalities is an untrivial task on a system as configurable as Mma.
@halirutan I don't believe that's going to weed out bugs on unreleased software. :I
@Szabolcs I must say that fixation by WRI on "knowledge" and being hip (aka Twitter oneliners) has negative effects on actual power user perceptions on them.
@kirma Looking back, you can really say that every first release of a new major version had a lot of hardcore bugs. When I remember right, one exception might be version 7, but don't some of you remember version 6.0. Completely useless. In version 9 there was such a hardcore bug, that users on windows (when I remember right) weren't even able to install Mathematica.
Most of that curated data is either not really useful, or the API implementations are actually too flaky to depend on them. And nobody on their right mind cares of being capable of writing a snippet of code less than 150 characters, or whatnot.
@halirutan I must admit I wonder how QA process at WRI works, if there's any. Then again, I sort of know that it's quite easy to make it fail, and CEO owning the company probably doesn't help if he doesn't grasp that priority.
It could be a freewheeling eighties methodology: we'll do something and when it seems to work, release it. Just that what works and makes sense for eighties word processors doesn't make sense for any large software system these days... and doesn't make sense in so many different levels.
Those new functions that were completely broken on v10.1 (despite having examples on the documentation) implied something about basic test coverage...
I think the best way to convince people to start using Mathematica is to make sure that within a group of coworkers, Mathematica-results are admired. "How did you calculate that so quickly?" "How did you manage to make that special custom plot that looks so good?", etc. I do get that sometimes, and I can show "magic" to people while using Mathematica :)
Unfortunately sometimes I also have to show how I work around annoying bugs ...
@kirma Does anyone here use the new test framework in v10? I did use the previous one from Workbench a bit, but I had so many problems with the new one, even in v10.1, that I just gave up...
@kirma Probably it's only us not understanding how this new twitter/facebook world really works and being annoyed by it, but somehow I have to get my work done.
I once made some gallup poll meta-analysis fitting with confidence intervals ahead of national elections, just as a hobby. I shared them in a bar with couple guys... some weeks later one told that head of one of large media houses in the country had been shown them, and found them very intriguing.
I think that once I had teached myself background on the field, the actual code producing those quite nice plots was like a half screen long (and mostly boilerplate stuff).
@halirutan Actually I only used MUnit seriously for MATLink. The setup we used there simply doesn't work with the MUnit from v10 (but it does work with Workbench.,
Then we have .mt files which may have a short setup at the beginning (such as loading the package, OpenMATLAB[], etc.). Then these are followed by tests, which are meant to be checked in order, as some of the evaluations depend on previous ones.
@Szabolcs But all this is different in the current Test-Framework. You essentially have only the possibility to set up a testnotebook containing VerificationTest
@Szabolcs I see. I'm a bit afraid to do some work at this when it is still so unfinished. I would have to decide a lot of design-questions by myself and in the worst case, everything is changed in some later versions of M
@Guesswhoitis. I noticed this when running examples from the help. I think the Benchmarking has a lot of LAPACK, so I was thinking in addition to the functionality tested by Benchmarking
@Guesswhoitis. i.e. I watch "top" command while running them, and a good number had the CPU > 100% implying multithreading. Maybe it's just my imagination
@Guesswhoitis. It was certainly possible, but the hash lookups for associations will improve performance for a large game, and the association["key","key"] syntax helps readability.
And I hadn't thought of how to start it yet. I made a puzzle game the old fashioned way. Then I wanted to look more at "knowledge network automata" as a way to avoid existing genre constraints. Then as I thought about how to implement and code such things this is where I ended up for now. Kind of half way between object oriented and functional/list oriented.
So now I add more entities and rules, and rules that only apply with certain constraints and conditions are met.
My goal is to start with a famous story, like Gilgamesh, and come up with a minimal set of entities, properties, and rules that allow the story to roughly play out automatically.
Then to make a game I would just add a set of player interactions and a goal to the system.
@Eric, the newest versions of BLAS/ATLAS and LAPACK are indeed multicore-adapted, IIRC. The problem is that I don't think there's a way to query Mathematica which build of those libraries is it using…
@Guesswhoitis. I suspect it works for "special regions", that is simple regions, in practice.
But some of failure modes with boolean regions are rather scary. Did they try them?
I started to have a nagging feeling about potential theoretical implementations they could have there... and sure, when I tried it out, it wasn't at all as pretty as one could hope.
But seriously... try RandomPoint@RegionDifference[Disk[], RegularPolygon[4]] ... once, or couple times. Even the error messages are garbled.
And that error message makes absolutely no sense.
Ah. It's probably mostly related to RegularPolygon, although it doesn't look awfully good in general either.
The Mathematica 10 documentation was updated for FindInstance adding support for regions.
In my use case, I'm trying to sample points in a set of disks:
region = DiscretizeRegion@RegionUnion@Table[Disk[RandomReal[4, {2}], RandomReal[1]], {10}]
FindInstance[{x, y} ∈ region, {x, y}, Reals, 2] // ...
@kirma I actually strongly disagree with this. The oneliner tweet PR stunts definitely speak to me. It shows people just how much Mma can do in so few keystrokes. I can't read the Neat Examples section for every function, so often the coolest oneliners introduce me to features I wouldn't have explored. I don't care that much about the process of shrinking something to fit in 140 characters.
while reg = DiscretizeRegion@BooleanRegion[Xor, {RegularPolygon[5], Disk[]}] takes a while, pulling points seems pretty fast to me RandomPoint[reg, 1000000]; // RepeatedTiming
@kirma Sure, there should be a balance. But how many new customers (i.e. $) does WR get by focusing x amount of effort on power users vs. people intrigued by oneliners
@Guesswhoitis. I bet random sampling from simplices is vastly easier than from any constant region. But those simplex-based methods... well, they work.
I just expected RandomPoint to work for at least some simple constructed non-mesh regions I try at it. It failed on the first one.
@kirma " I must say that fixation by WRI on "knowledge" and being hip (aka Twitter oneliners) has negative effects on actual power user perception" ... brilliant
@Guesswhoitis. As far as I've understood, "special regions" (I'd say more like elementary regions) probably have special-purpose implementations. When you start to derive regions using CSG from them, they're turned into general (in)equalities.
Mma is a tool more apt for scientists than for engineers. That precludes the help from some important cash sources. Now, if you add the known limitation for using it in anything but modelling and toy apps you get a financial nightmare for a company that size.
I mean... most primitives people play with, unless they go straight to discretization, are probably semialgebraic, and their combinations are also semialgebraic.
I have occassionally considered building semialgebaraic set aware discretization and sampling methods, but... hell, I'm not paid for it.
@Guesswhoitis. I believe the main problem is getting too big in a small market. If you dream with a large "the world is inside me" kind of project you have to have continuously growing monetary resources. That means expanding (or milking, to some extend) your market more and more every day.
@kirma I've read somewhere that the actual reason for humans to start cultivating cereals wasn't flour but beer. Sounds plausible
If I remind it well, the hypotheses was based in that archeologically identified fermentation recipients predates everything that could've been used as a mill
But seriously, cuneiform business contracts and disputes are interesting stuff to read. Those are a bit hard to find though, mostly written before the time of World Wide Web. :o
Literally, the very first command I evaluated in 10.2 $Failed although it worked since version 8!
And as it turns out, someone silently changed $UserDocumentsDirectory from being the home-path to ~/Documents under Linux. And of course, the documentation states that is should be ~ :-)
Thanks to this question I became aware of the fact that we have (at least) these overlapping tags:
probability
probability-and-stats
statistics
Are "probability" and "statistics" sufficiently differentiated to keep each of those tags and eliminate probability-and-stats or should all of these ...
I flagged this answer because I mis-read it and thought it was a thank-you answer and not a real one. I know my flag will be rejected and rightly so, but I wish I could have retracted it myself. The close-vote pop-up allows retraction, I request the flagging pop-up be given the same feature.
Is there a compound that absorbs all visible light?
